Performance comparison of remote procedure calling and mobile agent approach to control and data transfer in distributed computing environment

Abstract

Networks and distributed processing systems are of critical and growing importance in business, government and other organization. In this new millennium, it is increasingly important to retrieve and transfer information efficiently and rapidly from widely dispersed users in an enterprise network. This paper presents the use of mobile agent (MA) in the control and transfer of data in distributed computing environment as against the traditional client–server computing. The conceptual design of the MA is process-based and packet switching oriented. The mobility infrastructure was developed to facilitate a transmission control protocol/internet protocol socket-based connection between source and destination machine using agent transfer. To specify the path of itinerary agent, a model of dynamic route decisions was defined and implemented using concept of oldness vector and random number generator. In order to show a proof of a superior scheme provided by mobile agent over the traditional scheme, an analytical model was developed. From the analytical model, the following parameters were determined: bandwidth usage against number of requests per service, percentage denial of service measured against number of requests per service, percentage denial of service measured against network failure rate, and network bandwidth overload with retransmission. The simulated network has a state that is modeled after Bernoulli random variable (BRV) with probability of success p=80% and probability of failure q=20%. The network breakdown was implemented also by an event generated by a multiplicative congruential pseudorandom number generator. The computer programs for the mobile agent and the simulated model were written in C++ and Java programming languages. Experimental results based on the simulation were presented.